Latent variable modeling of complex high-dimensional data

复杂高维数据的潜变量建模

• 批准号: RGPIN-2019-05915
• 负责人: PedrosoEstevamdeSouza, Camila
• 金额: $ 1.17万
• 依托单位: University of Western Ontario
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=738592
• 关键词: Latent; variable; modeling; complex; dimensional

As an Assistant Professor in Data Science my research work consists of developing novel statistical methods to analyze large complex data structures so that important scientific questions can be answered. In particular, my proposed research program is focused on the development of techniques to analyze data when crucial information is absent. While my methods can be applied to a wide range of areas in the Natural Sciences and Engineering, I will consider two main application themes over the next five years: 1) analysis of electrical load data from substations and 2) DNA and RNA sequencing data from single cells. In Theme 1, I will consider aggregated energy consumption data measured over time at substations that serve a fixed number of consumers from different types (e.g., residential, commercial and industrial). The individual consumer-level energy usage curves are not observed, only the sums of individuals' energy usage. My goal is to develop novel statistical methods to infer the typical weekly energy usage curve for each type of consumer using the substation aggregated data and additional information, such as temperature and characteristics of the substations (e.g., capacity and number of low voltage feeders). In addition, I will cluster substations into different groups according to their consumer type-specific energy usage curves. This work will allow power companies around the world to better understand energy usage in order to provide adequate energy at low cost. In Theme 2, I will analyze DNA and RNA sequencing data obtained from individual cells. Because the amount of DNA/RNA material per cell is limited, the resulting single-cell sequencing data contain technical noise and a large amount of missing information. My goal is to build new statistical tools to infer the different groups of cells comprising a tissue based on their DNA/RNA sequencing data taking into account the challenges arising from this type of technology. My work on single-cell genomics will provide scientists in various areas of biology with the adequate set of statistical tools to assess the genomic composition of cells, which will lead to a better understanding of how individual cells differentiate to form tissues and how tissues work. My proposed research team consists of two MSc and two PhD students working under Theme 1 and one MSc and two PhD students working under Theme 2. All the resulting research work under Themes 1 and 2 will be submitted for publication at high-impact scientific journals, as well as presented by the students and myself at relevant conferences. In addition, all statistical methods developed will be implemented in the free software environment R and will include user-friendly tutorial guides.

作为数据科学助理教授，我的研究工作包括开发新颖的统计方法来分析大型复杂数据结构，以便回答重要的科学问题。特别是，我提出的研究计划侧重于开发在关键信息缺失时分析数据的技术。虽然我的方法可以应用于自然科学和工程的广泛领域，但我将在未来五年考虑两个主要应用主题：1）分析变电站的电力负荷数据，2）来自变电站的 DNA 和 RNA 测序数据单细胞。在主题 1 中，我将考虑在为不同类型（例如住宅、商业和工业）的固定数量消费者提供服务的变电站随时间测量的汇总能源消耗数据。不观察个人消费者层面的能源使用曲线，仅观察个人能源使用的总和。我的目标是开发新颖的统计方法，使用变电站汇总数据和附加信息（例如变电站的温度和特性（例如低压馈线的容量和数量））推断每种类型消费者的典型每周能源使用曲线。此外，我将根据特定消费者类型的能源使用曲线将变电站分为不同的组。这项工作将使世界各地的电力公司更好地了解能源使用情况，以便以低成本提供充足的能源。在主题 2 中，我将分析从单个细胞获得的 DNA 和 RNA 测序数据。由于每个细胞的 DNA/RNA 材料量有限，因此所得的单细胞测序数据包含技术噪声和大量缺失信息。我的目标是构建新的统计工具，根据 DNA/RNA 测序数据推断构成组织的不同细胞群，同时考虑到此类技术带来的挑战。我在单细胞基因组学方面的工作将为生物学各个领域的科学家提供一套足够的统计工具来评估细胞的基因组组成，这将有助于更好地了解单个细胞如何分化形成组织以及组织如何发挥作用。我提议的研究团队由在主题 1 下工作的两名硕士生和两名博士生以及在主题 2 下工作的一名硕士生和两名博士生组成。主题 1 和 2 下的所有研究成果将提交在高影响力的科学期刊上发表，以及学生和我自己在相关会议上的介绍。此外，开发的所有统计方法都将在自由软件环境 R 中实现，并将包括用户友好的教程指南。